Achieving Comprehensive Patient Care Using Natural Language Processing

Telehealth has seen unprecedented growth within the past few years. As a result of this adoption, at the onset of the COVID-19 pandemic, health care organizations were quickly able to adapt their processes from being rooted in in-person care to relying on virtual care.

But while health care technology companies continue to provide insights to health care stakeholders from structured clinical patient data stemming from telehealth visits, it is still a challenge to provide insights from the unstructured information that is received directly from patients. Further, access to telemedicine may be particularly challenging for low-income patients and patients in rural areas, who may not have reliable broadband, smartphones, or computers. These gaps in insight and access have a critical impact on overall patient well-being.

Real-world evidence companies still face challenges to extracting meaningful data from the unstructured clinical diagnosis and procedural notes derived from patient and provider conversations. This prevents health care stakeholders from leveraging this nonclinical patient information and social determinants of health (SDoH) data that are gathered during the virtual patient visit. Reinforcing this barrier, the usage of SDoH patient data was also recognized and acknowledged by the Office of the National Coordinator for Health Information Technology’s (ONC) Interoperability Standards Advisory in 2020 as a key challenge.

As the adoption of telehealth continues to grow in the future, the problem of generating meaningful insights from these semistructured and unstructured clinical notes and barriers to sharing insights with health care stakeholders restrict their ability to inform comprehensive patient care.

The high-level diagram below illustrates how real-world evidence companies can capture and leverage clinical notes and SDoH patient data using natural language processing (NLP) to generate insights from comprehensive patient data.

Using this approach, real-world evidence companies can scale their existing data platform—which currently processes the structured Health Level 7 (HL7), Fast Health care Interoperability Resources (FHIR), and Batch messages—to also process the semistructured and unstructured clinical notes and SDoH patient data to generate powerful data insights.

These data can go through the NLP interpreter for preprocessing. The NLP interpreter will extract and transform these data from various sources and convert it into a standard format. Converting the data into a standard format reduces the complexity of maintaining the data quality and integrity.

The transformed data is then sent to the NLP model optimizer to continuously train and optimize the data model, creating future efficiencies. These data are then sent to the NLP data model for generating meaningful insights that can be sent to the existing data platform.

These insights of a patient's health are accurate, complete, and consistent. This provides clinicians, hospitals, and various other health care stakeholders with a 360-degree longitudinal view of patient health.

As semistructured and unstructured clinical notes and SDoH patient data greatly impact a patient’s health, it's critical for real-world evidence companies to ensure this information is captured and shared with providers, hospitals, and other health care stakeholders for improved patient treatment outcomes. NLP in health care can revolutionize the way patient care is delivered by acting as a catalyst in the shift from caring for patients based on limited, available clinical information to creating a customized plan of comprehensive patient care, informed by a broader understanding of patient health.